import contextlib
import copy
import hashlib
import logging
import threading
import django.apps
from django.conf import settings
from django.db import connections
from django.db import DEFAULT_DB_ALIAS
from django.db import models
from django.db import router
from django.db import transaction
from django.db.models.expressions import Col
from django.db.models.fields.related import RelatedField
from django.db.models.sql import Query
from django.db.models.sql.datastructures import BaseTable
from django.db.utils import ProgrammingError
import pgconnection
LOGGER = logging.getLogger('pgtrigger')
_unset = object()
# Postgres only allows identifiers to be 63 chars max. Since "pgtrigger_"
# is the prefix for trigger names, and since an additional "_" and
# 5 character hash is added, the user-defined name of the trigger can only
# be 47 chars.
# NOTE: We can do something more sophisticated later by allowing users
# to name their triggers and then hashing the names when actually creating
# the triggers.
MAX_NAME_LENGTH = 47
# Installation states for a triggers
INSTALLED = 'INSTALLED'
UNINSTALLED = 'UNINSTALLED'
OUTDATED = 'OUTDATED'
PRUNE = 'PRUNE'
# All registered triggers for each model
registry = {}
# All triggers currently being ignored
_ignore = threading.local()
def _quote(label):
"""Conditionally wraps a label in quotes"""
if label.startswith('"'):
return label
else:
return f'"{label}"'
def _get_database(model):
"""
Obtains the database used for a trigger / model pair. The database
for the connection is selected based on the write DB in the database
router config.
"""
return router.db_for_write(model) or DEFAULT_DB_ALIAS
def _postgres_databases(databases):
"""Given an iterable of databases, only return postgres ones"""
return [database for database in databases if connections[database].vendor == 'postgresql']
def _get_connection(model):
"""
Obtains the connection used for a trigger / model pair. The database
for the connection is selected based on the write DB in the database
router config.
"""
return connections[_get_database(model)]
def _get_model(table):
"""Obtains a django model based on its table name"""
for model in django.apps.apps.get_models(): # pragma: no branch
if _quote(model._meta.db_table) == _quote(table):
return model
def _is_concurrent_statement(sql):
"""
True if the sql statement is concurrent and cannot be ran in a transaction
"""
sql = sql.strip().lower() if sql else ''
return sql.startswith('create') and 'concurrently' in sql
def _inject_pgtrigger_ignore(sql, sql_vars, cursor): # pragma: no cover
"""
A pgconnection pre_execute hook that sets a pgtrigger.ignore
variable in the executed SQL. This lets other triggers know when
they should ignore execution
"""
if cursor.name:
# A named cursor automatically prepends
# "NO SCROLL CURSOR WITHOUT HOLD FOR" to the query, which
# causes invalid SQL to be generated. There is no way
# to override this behavior in psycopg2, so context tracking
# is ignored for named cursors. Django only names cursors
# for iterators and other statements that read the database,
# so it seems to be safe to ignore named cursors.
# TODO(@wesleykendall): Find a way to generate valid SQL
# for local variables within a named cursor declaration.
return None
elif _is_concurrent_statement(sql):
# Concurrent index creation is incompatible with local variable
# setting. Ignore this specific statement for now
return None
sql = ('SET LOCAL pgtrigger.ignore=\'{' + ','.join(_ignore.value) + '}\';') + sql
return sql, sql_vars
[docs]def register(*triggers):
"""
Register the given triggers with wrapped Model class
"""
def _model_wrapper(model_class):
for trigger in triggers:
trigger.register(model_class)
return model_class
return _model_wrapper
class _Level:
def __init__(self, name):
self.name = name
def __str__(self):
return self.name
#: For specifying row-level triggers (the default)
Row = _Level('ROW')
#: For specifying statement-level triggers
Statement = _Level('STATEMENT')
[docs]class Referencing:
"""For specifying the REFERENCING construct of a statement-level trigger"""
def __init__(self, *, old=None, new=None):
if not old and not new:
raise ValueError(
'Must provide either "old" and/or "new" to the referencing'
' construct of a trigger'
)
self.old = old
self.new = new
def __str__(self):
ref = 'REFERENCING'
if self.old:
ref += f' OLD TABLE AS {self.old} '
if self.new:
ref += f' NEW TABLE AS {self.new} '
return ref
class _When:
def __init__(self, name):
self.name = name
def __str__(self):
return self.name
#: For specifying "BEFORE" in the "when" clause of a trigger
Before = _When('BEFORE')
#: For specifying "AFTER" in the "when" clause of a trigger
After = _When('AFTER')
#: For specifying "INSTEAD OF" in the "when" clause of a trigger
InsteadOf = _When('INSTEAD OF')
class _Operation:
def __init__(self, name):
self.name = name
def __or__(self, other):
return _Operations(self, other)
def __str__(self):
return self.name
class _Operations(_Operation):
def __init__(self, *operations):
self.operations = operations
def __str__(self):
return ' OR '.join(str(operation) for operation in self.operations)
#: For specifying "UPDATE" in the "operation" clause of a trigger
Update = _Operation('UPDATE')
#: For specifying "DELETE" in the "operation" clause of a trigger
Delete = _Operation('DELETE')
#: For specifying "TRUNCATE" in the "operation" clause of a trigger
Truncate = _Operation('TRUNCATE')
#: For specifying "INSERT" in the "operation" clause of a trigger
Insert = _Operation('INSERT')
[docs]class UpdateOf(_Operation):
"""For specifying "UPDATE OF" in the "operation" clause of a trigger"""
def __init__(self, *columns):
if not columns:
raise ValueError('Must provide at least one column')
self.columns = ', '.join(f'{_quote(col)}' for col in columns)
def __str__(self):
return f'UPDATE OF {self.columns}'
[docs]class Condition:
"""For specifying free-form SQL in the "condition" clause of a trigger"""
sql = None
def __init__(self, sql=None):
self.sql = sql or self.sql
if not self.sql:
raise ValueError('Must provide SQL to condition')
def resolve(self, model):
return self.sql
class _OldNewQuery(Query):
"""
A special Query object for referencing the OLD and NEW variables in a
trigger. Only used by the Q object
"""
def build_lookup(self, lookups, lhs, rhs):
# Django does not allow custom lookups on foreign keys, even though
# DISTINCT FROM is a comnpletely valid lookup. Trick django into
# being able to apply this lookup to related fields.
if lookups == ['df'] and isinstance(lhs.output_field, RelatedField):
lhs = copy.deepcopy(lhs)
lhs.output_field = models.IntegerField(null=lhs.output_field.null)
return super().build_lookup(lookups, lhs, rhs)
def build_filter(self, filter_expr, *args, **kwargs):
if isinstance(filter_expr, Q):
return super().build_filter(filter_expr, *args, **kwargs)
if filter_expr[0].startswith('old__'):
alias = 'OLD'
elif filter_expr[0].startswith('new__'):
alias = 'NEW'
else: # pragma: no cover
raise ValueError('Filter expression on trigger.Q object must reference old__ or new__')
filter_expr = (filter_expr[0][5:], filter_expr[1])
node, _ = super().build_filter(filter_expr, *args, **kwargs)
self.alias_map[alias] = BaseTable(alias, alias)
for child in node.children:
child.lhs = Col(
alias=alias,
target=child.lhs.target,
output_field=child.lhs.output_field,
)
return node, {alias}
[docs]class F(models.F):
"""
Similar to Django's F object, allows referencing the old and new
rows in a trigger condition.
"""
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
if self.name.startswith('old__'):
self.row_alias = 'OLD'
elif self.name.startswith('new__'):
self.row_alias = 'NEW'
else:
raise ValueError('F() values must reference old__ or new__')
self.col_name = self.name[5:]
@property
def resolved_name(self):
return f'{self.row_alias}.{_quote(self.col_name)}'
def resolve_expression(self, query=None, *args, **kwargs):
return Col(
alias=self.row_alias,
target=query.model._meta.get_field(self.col_name),
)
[docs]@models.fields.Field.register_lookup
class IsDistinctFrom(models.Lookup):
"""
A custom IS DISTINCT FROM field lookup for common trigger conditions
"""
lookup_name = 'df'
def as_sql(self, compiler, connection):
lhs, lhs_params = self.process_lhs(compiler, connection)
rhs, rhs_params = self.process_rhs(compiler, connection)
params = lhs_params + rhs_params
return '%s IS DISTINCT FROM %s' % (lhs, rhs), params
[docs]@models.fields.Field.register_lookup
class IsNotDistinctFrom(models.Lookup):
"""
A custom IS NOT DISTINCT FROM field lookup for common trigger conditions
"""
lookup_name = 'ndf'
def as_sql(self, compiler, connection):
lhs, lhs_params = self.process_lhs(compiler, connection)
rhs, rhs_params = self.process_rhs(compiler, connection)
params = lhs_params + rhs_params
return '%s IS NOT DISTINCT FROM %s' % (lhs, rhs), params
[docs]class Q(models.Q, Condition):
"""
Similar to Django's Q object, allows referencing the old and new
rows in a trigger condition.
"""
def resolve(self, model):
connection = _get_connection(model)
query = _OldNewQuery(model)
sql = (
connection.cursor()
.mogrify(
*self.resolve_expression(query).as_sql(
compiler=query.get_compiler('default'),
connection=connection,
)
)
.decode()
.replace('"OLD"', 'OLD')
.replace('"NEW"', 'NEW')
)
return sql
def _drop_trigger(table, trigger_pgid):
model = _get_model(table)
connection = _get_connection(model)
with connection.cursor() as cursor:
cursor.execute(f'DROP TRIGGER IF EXISTS {trigger_pgid} ON {_quote(table)};')
# Allows Trigger methods to be used as context managers, mostly for
# testing purposes
@contextlib.contextmanager
def _cleanup_on_exit(cleanup):
yield
cleanup()
[docs]class Trigger:
"""
For specifying a free-form PL/pgSQL trigger function or for
creating derived trigger classes.
"""
name = None
level = Row
when = None
operation = None
condition = None
referencing = None
func = None
declare = None
def __init__(
self,
*,
name=None,
level=None,
when=None,
operation=None,
condition=None,
referencing=None,
func=None,
declare=None,
):
self.name = name or self.name
self.level = level or self.level
self.when = when or self.when
self.operation = operation or self.operation
self.condition = condition or self.condition
self.referencing = referencing or self.referencing
self.func = func or self.func
self.declare = declare or self.declare
if not self.level or not isinstance(self.level, _Level):
raise ValueError(f'Invalid "level" attribute: {self.level}')
if not self.when or not isinstance(self.when, _When):
raise ValueError(f'Invalid "when" attribute: {self.when}')
if not self.operation or not isinstance(self.operation, _Operation):
raise ValueError(f'Invalid "operation" attribute: {self.operation}')
if self.level == Row and self.referencing:
raise ValueError('Row-level triggers cannot have a "referencing" attribute')
if not self.name:
raise ValueError('Trigger must have "name" attribute')
self.validate_name()
def __str__(self):
return self.name
def validate_name(self):
"""Verifies the name is under the maximum length"""
if len(self.name) > MAX_NAME_LENGTH:
raise ValueError(f'Trigger name "{self.name}" > {MAX_NAME_LENGTH} characters.')
def get_pgid(self, model):
"""The ID of the trigger and function object in postgres
All objects are prefixed with "pgtrigger_" in order to be
discovered/managed by django-pgtrigger
"""
model_hash = hashlib.sha1(self.get_uri(model).encode()).hexdigest()[:5]
pgid = f'pgtrigger_{self.name}_{model_hash}'
if len(pgid) > 63:
raise ValueError(f'Trigger identifier "{pgid}" is greater than 63 chars')
# NOTE - Postgres always stores names in lowercase. Ensure that all
# generated IDs are lowercase so that we can properly do installation
# and pruning tasks.
return pgid.lower()
def get_condition(self, model):
return self.condition
def get_declare(self, model):
"""
Gets the DECLARE part of the trigger function if any variables
are used.
Returns:
List[tuple]: A list of variable name / type tuples that will
be shown in the DECLARE. For example [('row_data', 'JSONB')]
"""
return self.declare or []
def get_func(self, model):
"""
Returns the trigger function that comes between the BEGIN and END
clause
"""
if not self.func:
raise ValueError('Must define func attribute or implement get_func')
return self.func
def get_uri(self, model):
"""The URI for the trigger in the registry"""
return f'{model._meta.app_label}.{model._meta.object_name}:{self.name}'
def register(self, *models):
"""Register model classes with the trigger"""
# Compute the unique trigger function names that are already
# registered in order to prevent an accidental collision
registered_function_names = {
trigger.get_pgid(model) for model, trigger in registry.values()
}
for model in models:
uri = self.get_uri(model)
if uri in registry:
raise ValueError(
f'Trigger with name "{self.name}" is already'
f' registered for model "{model}"'
)
if self.get_pgid(model) in registered_function_names:
raise ValueError(
f'Trigger with name "{self.name}" on model "{model}"'
' has a trigger function name that is already taken.'
' Use a different name for the trigger.'
)
registry[self.get_uri(model)] = (model, self)
return _cleanup_on_exit(lambda: self.unregister(*models))
def unregister(self, *models):
"""Unregister model classes with the trigger"""
for model in models:
del registry[self.get_uri(model)]
return _cleanup_on_exit(lambda: self.register(*models))
def render_condition(self, model):
"""Renders the condition SQL in the trigger declaration"""
condition = self.get_condition(model)
resolved = condition.resolve(model).strip() if condition else ''
if resolved:
if not resolved.startswith('('):
resolved = f'({resolved})'
resolved = f'WHEN {resolved}'
return resolved
def render_declare(self, model):
"""Renders the DECLARE of the trigger function, if any"""
declare = self.get_declare(model)
if declare:
rendered_declare = 'DECLARE \n' + '\n'.join(
f'{var_name} {var_type};' for var_name, var_type in declare
)
else:
rendered_declare = ''
return rendered_declare
def render_ignore(self, model):
"""
Renders the clause that can dynamically ignore the trigger's execution
"""
return '''
IF (_pgtrigger_should_ignore(TG_TABLE_NAME, TG_NAME) IS TRUE) THEN
IF (TG_OP = 'DELETE') THEN
RETURN OLD;
ELSE
RETURN NEW;
END IF;
END IF;
'''
def render_func(self, model):
"""Renders the trigger function SQL statement"""
return f'''
CREATE OR REPLACE FUNCTION {self.get_pgid(model)}()
RETURNS TRIGGER AS $$
{self.render_declare(model)}
BEGIN
{self.render_ignore(model)}
{self.get_func(model)}
END;
$$ LANGUAGE plpgsql;
'''
def render_trigger(self, model):
"""Renders the trigger declaration SQL statement"""
table = model._meta.db_table
pgid = self.get_pgid(model)
return f'''
DO $$ BEGIN
CREATE TRIGGER {pgid}
{self.when} {self.operation} ON {_quote(table)}
{self.referencing or ''}
FOR EACH {self.level} {self.render_condition(model)}
EXECUTE PROCEDURE {pgid}();
EXCEPTION
-- Ignore issues if the trigger already exists
WHEN duplicate_object THEN null;
END $$;
'''
def render_comment(self, model):
"""Renders the trigger commment SQL statement
pgtrigger comments the hash of the trigger in order for us to
determine if the trigger definition has changed
"""
pgid = self.get_pgid(model)
hash = self.get_hash(model)
table = model._meta.db_table
return f'COMMENT ON TRIGGER {pgid} ON {_quote(table)} IS \'{hash}\''
def get_installation_status(self, model):
"""Returns the installation status of a trigger.
The return type is (status, enabled), where status is one of:
1. ``INSTALLED``: If the trigger is installed
2. ``UNINSTALLED``: If the trigger is not installed
3. ``OUTDATED``: If the trigger is installed but
has been modified
"enabled" is True if the trigger is installed and enabled or false
if installed and disabled (or uninstalled).
"""
connection = _get_connection(model)
trigger_exists_sql = f'''
SELECT oid, obj_description(oid) AS hash, tgenabled AS enabled
FROM pg_trigger
WHERE tgname='{self.get_pgid(model)}'
AND tgrelid='{model._meta.db_table}'::regclass;
'''
try:
with connection.cursor() as cursor:
cursor.execute(trigger_exists_sql)
results = cursor.fetchall()
except ProgrammingError: # pragma: no cover
# When the table doesn't exist yet, possibly because migrations
# haven't been executed, a ProgrammingError will happen because
# of an invalid regclass cast. Return 'UNINSTALLED' for this
# case
return (UNINSTALLED, None)
if not results:
return (UNINSTALLED, None)
else:
hash = self.get_hash(model)
if hash != results[0][1]:
return (OUTDATED, results[0][2] == 'O')
else:
return (INSTALLED, results[0][2] == 'O')
def get_hash(self, model):
"""
Computes a hash for the trigger, which is used to
uniquely identify its contents. The hash is computed based
on the trigger function and declaration.
Note: If the trigger definition includes dynamic data, such
as the current time, the trigger hash will always change and
appear to be out of sync.
"""
rendered_func = self.render_func(model)
rendered_trigger = self.render_trigger(model)
return hashlib.sha1(f'{rendered_func} {rendered_trigger}'.encode()).hexdigest()
def install(self, model):
"""Installs the trigger for a model"""
connection = _get_connection(model)
# Ensure we have the function to ignore execution of triggers
install_ignore_func(database=_get_database(model))
rendered_func = self.render_func(model)
rendered_trigger = self.render_trigger(model)
rendered_comment = self.render_comment(model)
with connection.cursor() as cursor:
cursor.execute(rendered_func)
cursor.execute(rendered_trigger)
cursor.execute(rendered_comment)
return _cleanup_on_exit(lambda: self.uninstall(model))
def uninstall(self, model):
"""Uninstalls the trigger for a model"""
_drop_trigger(model._meta.db_table, self.get_pgid(model))
return _cleanup_on_exit(lambda: self.install(model)) # pragma: no branch
def enable(self, model):
"""Enables the trigger for a model"""
connection = _get_connection(model)
with connection.cursor() as cursor:
cursor.execute(
f'ALTER TABLE {_quote(model._meta.db_table)}'
f' ENABLE TRIGGER {self.get_pgid(model)};'
)
return _cleanup_on_exit(lambda: self.disable(model)) # pragma: no branch
def disable(self, model):
"""Disables the trigger for a model"""
connection = _get_connection(model)
with connection.cursor() as cursor:
cursor.execute(
f'ALTER TABLE {_quote(model._meta.db_table)}'
f' DISABLE TRIGGER {self.get_pgid(model)};'
)
return _cleanup_on_exit(lambda: self.enable(model)) # pragma: no branch
@contextlib.contextmanager
def ignore(self, model):
"""Ignores the trigger in a single thread of execution"""
with contextlib.ExitStack() as pre_execute_hook:
# Create the table name / trigger name URI to pass down to the
# trigger.
ignore_uri = f'{model._meta.db_table}:{self.get_pgid(model)}'
if not hasattr(_ignore, 'value'):
_ignore.value = set()
if not _ignore.value:
# If this is the first time we are ignoring trigger execution,
# register the pre_execute_hook
pre_execute_hook.enter_context(
pgconnection.pre_execute_hook(_inject_pgtrigger_ignore)
)
if ignore_uri not in _ignore.value:
try:
_ignore.value.add(ignore_uri)
yield
finally:
_ignore.value.remove(ignore_uri)
else: # The trigger is already being ignored
yield
if not _ignore.value and transaction.get_connection().in_atomic_block:
# We've finished all ignoring of triggers, but we are in a transaction
# and still have a reference to the local variable. Reset it
with transaction.get_connection().cursor() as cursor:
cursor.execute('RESET pgtrigger.ignore;')
[docs]class Protect(Trigger):
"""A trigger that raises an exception"""
when = Before
def get_func(self, model):
return f'''
RAISE EXCEPTION
'pgtrigger: Cannot {str(self.operation).lower()} rows from % table',
TG_TABLE_NAME;
'''
[docs]class FSM(Trigger):
"""Enforces a finite state machine on a field.
Supply the trigger with the "field" that transitions and then
a list of tuples of valid transitions to the "transitions" argument.
.. note::
Only non-null ``CharField`` fields are currently supported.
"""
when = Before
operation = Update
field = None
transitions = None
def __init__(self, *, name=None, condition=None, field=None, transitions=None):
self.field = field or self.field
self.transitions = transitions or self.transitions
if not self.field: # pragma: no cover
raise ValueError('Must provide "field" for FSM')
if not self.transitions: # pragma: no cover
raise ValueError('Must provide "transitions" for FSM')
super().__init__(name=name, condition=condition)
def get_declare(self, model):
return [('_is_valid_transition', 'BOOLEAN')]
def get_func(self, model):
col = model._meta.get_field(self.field).column
transition_uris = '{' + ','.join([f'{old}:{new}' for old, new in self.transitions]) + '}'
return f'''
SELECT CONCAT(OLD.{_quote(col)}, ':', NEW.{_quote(col)}) = ANY('{transition_uris}'::text[])
INTO _is_valid_transition;
IF (_is_valid_transition IS FALSE AND OLD.{_quote(col)} IS DISTINCT FROM NEW.{_quote(col)}) THEN
RAISE EXCEPTION
'pgtrigger: Invalid transition of field "{self.field}" from "%" to "%" on table %',
OLD.{_quote(col)},
NEW.{_quote(col)},
TG_TABLE_NAME;
ELSE
RETURN NEW;
END IF;
''' # noqa
[docs]class SoftDelete(Trigger):
"""Sets a field to a value when a delete happens.
Supply the trigger with the "field" that will be set
upon deletion and the "value" to which it should be set.
The "value" defaults to False.
.. note::
This trigger currently only supports nullable ``BooleanField``,
``CharField``, and ``IntField`` fields.
"""
when = Before
operation = Delete
field = None
value = False
def __init__(self, *, name=None, condition=None, field=None, value=_unset):
self.field = field or self.field
self.value = value if value is not _unset else self.value
if not self.field: # pragma: no cover
raise ValueError('Must provide "field" for soft delete')
super().__init__(name=name, condition=condition)
def get_func(self, model):
soft_field = model._meta.get_field(self.field).column
pk_col = model._meta.pk.column
def _render_value():
if self.value is None:
return 'NULL'
elif isinstance(self.value, str):
return f"'{self.value}'"
else:
return str(self.value)
return f'''
UPDATE {_quote(model._meta.db_table)}
SET {soft_field} = {_render_value()}
WHERE {_quote(pk_col)} = OLD.{_quote(pk_col)};
RETURN NULL;
'''
[docs]def get(*uris, database=None):
"""
Get triggers matching URIs or all triggers registered to models.
If a database is provided, will only enable triggers
registered to a particular database.
A URI is in the format of "app_label.model_name:trigger_name"
"""
if database and uris:
raise ValueError('Cannot supply both trigger URIs and a database')
if not database:
databases = {_get_database(model) for model, _ in registry.values()}
else:
databases = [database] if isinstance(database, str) else database
if uris:
for uri in uris:
if uri and len(uri.split(':')) == 1:
raise ValueError(
'Trigger URI must be in the format of "app_label.model_name:trigger_name"'
)
elif uri and uri not in registry:
raise ValueError(f'URI "{uri}" not found in pgtrigger registry')
return [registry[uri] for uri in uris]
else:
return [
(model, trigger)
for model, trigger in registry.values()
if _get_database(model) in databases
]
[docs]def install(*uris, database=None):
"""
Install registered triggers matching URIs or all triggers if URIs aren't
provided. If URIs aren't provided, prune any orphaned triggers from the
database. If a database is provided, will only install triggers
registered to a particular database.
"""
if uris:
model_triggers = get(*uris, database=database)
else:
model_triggers = [
(model, trigger)
for model, trigger in get(database=database)
if trigger.get_installation_status(model)[0] != INSTALLED
]
for model, trigger in model_triggers:
LOGGER.info(
f'pgtrigger: Installing {trigger} trigger'
f' for {model._meta.db_table} table'
f' on {_get_database(model)} database.'
)
trigger.install(model)
if not uris: # pragma: no branch
prune(database=database)
def get_prune_list(database=None):
"""Return triggers that will be pruned upon next full install
Args:
database (str, default=None): Only return results from this
database. Defaults to returning results from all databases
"""
installed = {
(_quote(model._meta.db_table), trigger.get_pgid(model)) for model, trigger in get()
}
if isinstance(database, str):
databases = [database]
else:
databases = database or settings.DATABASES
prune_list = []
for database in _postgres_databases(databases):
with connections[database].cursor() as cursor:
cursor.execute(
'SELECT tgrelid::regclass, tgname, tgenabled'
' FROM pg_trigger'
' WHERE tgname LIKE \'pgtrigger_%\''
)
triggers = set(cursor.fetchall())
prune_list += [
(trigger[0], trigger[1], trigger[2] == 'O', database)
for trigger in triggers
if (_quote(trigger[0]), trigger[1]) not in installed
]
return prune_list
[docs]def prune(database=None):
"""
Remove any pgtrigger triggers in the database that are not used by models.
I.e. if a model or trigger definition is deleted from a model, ensure
it is removed from the database
Args:
database (str, default=None): Only prune triggers from this
database. Defaults to returning results from all databases
"""
for trigger in get_prune_list(database=database):
LOGGER.info(
f'pgtrigger: Pruning trigger {trigger[1]}'
f' for table {trigger[0]} on {trigger[3]} database.'
)
_drop_trigger(trigger[0], trigger[1])
[docs]def enable(*uris, database=None):
"""
Enables registered triggers matching URIs or all triggers if no URIs
are provided. If a database is provided, will only enable triggers
registered to a particular database.
"""
if uris:
model_triggers = get(*uris, database=database)
else:
model_triggers = [
(model, trigger)
for model, trigger in get(database=database)
if trigger.get_installation_status(model)[1] is False
]
for model, trigger in model_triggers:
LOGGER.info(
f'pgtrigger: Enabling {trigger} trigger'
f' for {model._meta.db_table} table'
f' on {_get_database(model)} database.'
)
trigger.enable(model)
[docs]def uninstall(*uris, database=None):
"""
Uninstalls registered triggers matching URIs or all triggers if no
URIs are provided. If no URIs are provided, will also try to prune
any lingering triggers that are no longer in the code base.
Running migrations will re-install any existing triggers. This
behavior is overridable with ``settings.PGTRIGGER_INSTALL_ON_MIGRATE``
If a database is provided, will only uninstall triggers
registered to a particular database.
"""
if uris:
model_triggers = get(*uris, database=database)
else:
model_triggers = [
(model, trigger)
for model, trigger in get(database=database)
if trigger.get_installation_status(model)[0] != UNINSTALLED
]
for model, trigger in model_triggers:
LOGGER.info(
f'pgtrigger: Uninstalling {trigger} trigger'
f' for {model._meta.db_table} table'
f' on {_get_database(model)} database.'
)
trigger.uninstall(model)
if not uris:
prune(database=database)
[docs]def disable(*uris, database=None):
"""
Disables registered triggers matching URIs or all triggers if no URIs are
provided. If a database is provided, will only disable triggers
registered to a particular database.
"""
if uris:
model_triggers = get(*uris, database=database)
else:
model_triggers = [
(model, trigger)
for model, trigger in get(database=database)
if trigger.get_installation_status(model)[1]
]
for model, trigger in model_triggers:
LOGGER.info(
f'pgtrigger: Disabling {trigger} trigger for'
f' {model._meta.db_table} table'
f' on {_get_database(model)} database.'
)
trigger.disable(model)
[docs]@contextlib.contextmanager
def ignore(*uris):
"""
Dynamically ignore registered triggers matching URIs from executing in
an individual thread.
If no URIs are provided, ignore all pgtriggers from executing in an
individual thread.
"""
with contextlib.ExitStack() as stack:
for model, trigger in get(*uris):
stack.enter_context(trigger.ignore(model))
yield
def install_ignore_func(database=None):
"""
pgtrigger uses a special postgres function to determine when a trigger
should be ignored. This installs the function.
This function is automatically installed when all triggers are installed
with pgtrigger.install()
"""
databases = {_get_database(model) for model, _ in get(database=database)}
for database in _postgres_databases(databases):
with connections[database].cursor() as cursor:
cursor.execute(
'''
CREATE OR REPLACE FUNCTION _pgtrigger_should_ignore(
table_name NAME,
trigger_name NAME
)
RETURNS BOOLEAN AS $$
DECLARE
_pgtrigger_ignore TEXT[];
_result BOOLEAN;
BEGIN
BEGIN
SELECT INTO _pgtrigger_ignore
CURRENT_SETTING('pgtrigger.ignore');
EXCEPTION WHEN OTHERS THEN
END;
IF _pgtrigger_ignore IS NOT NULL THEN
SELECT CONCAT(table_name, ':', trigger_name) = ANY(_pgtrigger_ignore)
INTO _result;
RETURN _result;
ELSE
RETURN FALSE;
END IF;
END;
$$ LANGUAGE plpgsql;
'''
)